“Our AI is working great.”
That’s what a CMO told me three months after launch. Their dashboard looked incredible. 10,000 accounts enriched monthly. 50,000 personalized emails sent. 200+ automated workflows running.
Then I asked about pipeline.
Silence.
Sales was still complaining about lead quality. Win rates hadn’t moved. CAC was climbing. But hey, email open rates were up 3%.
This is where most AI initiatives die. Not from bad technology, but from measuring the wrong things and never closing the feedback loop.
If you’re just joining us, hiii. Welcome. You’re jumping in just in time to hear me finish out this rant on building an AI-based GTM infrastructure that’s rooted in reality, not that insane “I just made an agent that replaced my 9-person marketing team” nonsense.
A quick recap:
- Part 1: Your data is clean.
- Part 2: Your AI knows what qualified means.
- Part 3: Your workflows are automated.
- Part 4: Your content is modular and informed.
👋 Hiii, it’s Kaylee Edmondson and welcome to Looped In, the newsletter exploring demand gen and growth frameworks in B2B SaaS. I write this newsletter every Sunday (and occasionally on Mondays after a holiday), and wildly, a few thousand of you read it each week. I’m grateful!
Most Teams Are Measuring the Wrong Things
Last month I reviewed an AI dashboard. 47 different metrics. I asked which ones they used to make decisions.
The VP pointed to three: email open rates, form fills, and “AI efficiency score.”
But none of those tell you if AI is helping you close more deals.
The metrics that made marketing “provable” for the last decade don’t work anymore. Paid ads are 2x more expensive than three years ago. SEO traffic is moving into ChatGPT responses where you can’t track clicks. The direct response playbook is breaking down.
Meanwhile, AI lets marketing impact parts of the funnel we couldn’t really touch before. Sales qualification. Deal velocity. Rep productivity.
You need a different scorecard. One that tracks influence (the hard-to-measure stuff that’s becoming critical), demand (the traditional stuff that still matters), and revenue impact (the new territory marketing can finally claim).
Track Influence (Even Though It’s Messy)
I know, I know. These metrics feel squishy. But as paid efficiency becomes harder and harder, you need to know if your influence work is actually working.
Share of category conversation: When prospects research solutions in your space, how often does your brand show up?
This used to be SEO rankings. Now it’s also tracking whether you appear in AI-generated responses, analyst reports, practitioner content, community discussions.
One client started systematically tracking this quarterly. They found they were mentioned in 22% of relevant AI responses in Q1, 31% by Q2, 39% by Q3. That correlated directly with increased inbound volume and shorter sales cycles.
Branded search trends from target accounts: Track overall branded search volume, but more importantly, track which ICP companies are actually visiting your site.
A Series B client uses visitor identification to see which companies hit their site from branded search. They segment those visitors by ICP fit - company size, industry, tech stack.
They found that while total branded search traffic grew 11%, traffic from their core ICP (Series B-C SaaS, 100-500 employees) grew 34%. More importantly, those ICP visitors were 3x more likely to book meetings.
That told them their positioning work was reaching and resonating with the right audience, not just creating general awareness.
Target account engagement with thought leadership: Content engagement tracked by account tier in your CRM.
One demand gen team realized their viral LinkedIn posts got tons of engagement but almost none from Tier 1 accounts. They shifted to technical deep-dives. Total engagement dropped 40% but pipeline from content-influenced deals nearly tripled.
These metrics are imperfect. But when traditional performance marketing gets expensive, you need to know if your brand work is creating actual demand.
Track Demand (The Stuff That Still Matters)
Traditional demand gen metrics haven’t disappeared. But you need to track them differently.
Lead quality by source, not just lead volume: One client generated 2,400 leads quarterly. Only 380 met actual ICP criteria. They were wasting 80%+ of their budget on poor-fit leads.
They rebuilt targeting to focus on quality. Lead volume dropped to 1,100 but qualified leads increased to 720. Marketing became more efficient by doing less.
Meeting booking rate by signal type: Not all inbound is equal.
Tracked which behaviors predict meetings. Things like pricing page visits, vs technical documentation downloads, vs integration guides.
Shift your follow-up strategy to prioritize those high-conversion signals.
Channel efficiency as paid performance declines: Track cost per qualified opportunity, not cost per lead.
When one client’s paid CPC increased 60% year-over-year, they had to get ruthless about efficiency. They killed channels with high CPL but low meeting conversion. Reallocated budget to channels that drove actual pipeline even if CPL was higher.
The pattern: measure what actually predicts pipeline, not what’s easy to track.
Track Revenue Impact (The Newer Territory)
This is where it gets interesting. AI lets marketing finally prove impact on sales acceleration.
Deal velocity by engagement source: How long from opportunity to close, broken down by how the account was engaged?
One client’s numbers: Deals from cold outbound averaged 112 days to close. Inbound deals averaged 83 days. Deals from AI-triggered signal-based plays averaged 54 days.
Same product, same sales team, completely different velocity because of when and how they engaged.
Sales productivity multiplier: How many accounts can one rep effectively work when AI handles research and prep?
Pre-AI: One SDR client managed 40-50 active accounts. Post-AI with proper copilot tools: same SDR manages 95 accounts with higher connection rates because AI eliminated busy work.
That’s not replacing SDRs. That’s making each one 2x more effective.
Win rate improvement from better qualification: Your closed-won rate should improve as AI helps identify better-fit accounts earlier.
A demand gen team rebuilt their qualification around behavioral signals. Their win rate from qualified pipeline improved from 28% to 37% in six months. Better qualification at the front meant higher quality throughout.
Time saved per rep per week: When AI handles prospecting research, account prep, and follow-up drafting, how much time does sales get back?
One sales team tracked this religiously. Each rep reclaimed 5.8 hours weekly. That’s ~15% of their week freed up for actual selling instead of administrative work.
These aren’t traditional marketing metrics. But when AI lets you impact deal velocity and sales productivity, you need to track them. Otherwise you can’t prove the value you’re creating.
The Dashboard Nobody’s Building Yet
Truthfully, there are still just so many marketing team’s reporting out MQLs and email open rates when all your bosses care about is what marketing is doing to drive/impact revenue. Full stop.
The answer is hiding in metrics you’re not tracking yet.
Influence metrics prove your brand work is creating demand, not just impressions. Demand metrics show you’re generating qualified pipeline efficiently. Revenue metrics prove you’re helping close deals faster and at higher rates.
Break these down by account tier. Your Tier 1 accounts should show completely different performance than before AI.
Most teams won’t do this. They’ll keep tracking the old metrics while playing a new game.
You don’t have to be most teams.
The Discipline Nobody Wants to Hear About
Most marketing teams think in campaigns. AI requires you to think in systems.
Campaign thinking asks: “What campaign should we launch to hit our goal?”
Systems thinking asks: “Where is the breakdown in our engine, and what change will fix it?”
When demo requests drop, campaign thinking runs a LinkedIn ad campaign to drive more demos. Systems thinking asks whether it’s a volume problem (not enough traffic), a qualification problem (wrong visitors), or a conversion problem (qualified visitors not booking).
Those require completely different solutions. Volume needs more reach. Qualification needs better targeting. Conversion needs better landing page experience or offer.
AI amplifies whichever approach you take. Campaign thinking means AI helps you run more mediocre campaigns faster. Systems thinking means AI helps you build a better engine.
Structure Your Experiments Like You Mean It
Every initiative needs five components:
Hypothesis: Based on [data or pattern], we believe [specific action] will improve [specific metric] for [specific segment] because [rationale].
Not “let’s try competitive messaging.” Instead: “Accounts actively comparing us to competitors will respond better to direct feature comparisons because they’re already in evaluation mode, not education mode.”
Design: Treatment group, control group, success threshold, timeline, minimum sample size.
Execution: Run it without changing variables mid-test.
Analysis: What happened and why? Not just “did it work” but what you learned about your accounts, signals, or messaging.
Decision: Scale it, iterate it, or kill it. No “interesting insights we’ll think about.”
The discipline is in actually following this structure every time, not just for big initiatives.
Set Decision Triggers First
Before launching any experiment, complete this sentence:
“If this achieves [metric] within [timeframe] with [sample size], we will [action].”
If you can’t complete that sentence, you’re not ready to run the experiment. You’re just hoping something works.
The metric needs to be outcome-based (meetings booked, opportunities created, deals closed) not activity-based (emails sent, clicks generated).
The timeframe needs to match your sales cycle. B2B with 90-day cycles can’t evaluate experiments in 30 days.
The action needs to be specific. “Scale it” means what exactly? Add headcount? Increase budget? Automate the workflow?
Build Three Capabilities Your Team Doesn’t Have
Data literacy: Your marketers need to pull their own data, understand statistical significance, and interpret test results without waiting for RevOps.
This doesn’t mean everyone becomes a data analyst. It means everyone can answer basic questions: How many accounts in our Tier 1 showed high intent last month? What’s our demo booking rate from that segment? How does it compare to last quarter?
Technical systems knowledge: Your team needs to understand how data flows through your infrastructure. Not coding, but basic comprehension of where AI makes decisions, how signals trigger workflows, what happens when something breaks.
When a workflow isn’t firing correctly, your team should be able to diagnose whether it’s a data issue, a logic issue, or an integration issue. Not fix it themselves necessarily, but identify where the breakdown is.
Hypothesis-driven thinking: Stop launching campaigns to “increase awareness” or “generate engagement.”
Start with: “We believe [target segment] will [specific behavior] because [rationale based on data or insight].”
That shift forces you to define success upfront and think through why something should work before you execute it.
Who Actually Owns This
Nobody should “own” AI. That’s how you create silos.
AI needs distributed ownership with clear coordination:
RevOps owns infrastructure: Data architecture, system integration, signal capture, workflow automation, experiment tracking systems.
Demand Gen owns strategy: Campaign design, experiment execution, performance analysis, feedback from campaigns back to infrastructure.
Content owns production: Quality standards, prompt libraries, output QA, content performance analysis.
Sales owns execution: Following up on signals, providing structured feedback on lead quality, executing playbooks.
The coordination is what makes it work. Without it, you have four teams running their own AI strategies that don’t connect.
Coordination mechanisms that I use with my clients:
Weekly signal reviews: RevOps, demand gen, and sales review which signals are predicting outcomes and which need adjustment.
Bi-weekly experiment reviews: Demand gen and content review active experiments, decide what to scale or kill.
Monthly systems reviews: RevOps and demand gen review infrastructure performance, identify bottlenecks, discuss new capabilities needed.
Quarterly strategy planning: All functions together review learnings, adjust ICP and account tiers, plan big bets.
What creates expensive silos that nobody uses:
Creating an “AI Center of Excellence” separate from GTM teams. Making RevOps responsible for AI strategy on top of infrastructure maintenance. Having one “AI person” who owns all knowledge. Relying on consultants to run your AI indefinitely.
So many teams are doing this. Don’t do this. 👆
It Gets Better Over Time, If You Let It
AI GTM infrastructure doesn’t launch. It evolves.
Your assumptions about ICP will be wrong. Your initial signal weights will need adjustment. Your first playbooks will need iteration.
The first 90 days will feel so messy. You’ll have false positives—accounts flagged as high intent that aren’t. You’ll have false negatives—good accounts your scoring missed. Your workflows will need tweaking.
Most executives see that messiness and pull the plug. They wanted a solution, not a process.
But if you stick with it (and can convince your execs to stick with it, too):
Traditional marketing: Run campaign at X performance. Run another campaign at X performance. Linear results at best.
Systems approach: Start at X. Close feedback loops, improve to 1.2X. Iterate based on learnings, 1.5X. Add new playbooks, 2X. Compounding improvement.
The difference compounds over time. After six months, you’re 50% better. After 12 months, 100% better. After 18 months, 200% better—not from working harder but from the system getting smarter.
Your AI learns which signals actually predict pipeline. Your content gets optimized based on what moves deals forward. Your team kills what doesn’t work and doubles down on what does. Experiments become proven playbooks.
The competitive advantage isn’t the technology. Every company can buy the same tools, access the same data, use the same AI models.
The competitive advantage is organizational discipline.
Companies that win invested in infrastructure before buying AI tools. They taught their AI what “qualified” means for their specific business. They built proper workflows with human checkpoints at critical moments. They measured what actually matters and closed feedback loops religiously. They developed team capabilities systematically. They embraced continuous evolution over one-time launches.
None of that is sexy. Nobody posts about data hygiene or feedback loops on LinkedIn.
Most companies won’t do this work. They’ll skip the foundation in Parts 1-3 and jump straight to AI content generation in Part 4. They’ll treat it as a launch instead of a process. They’ll measure vanity metrics instead of revenue impact.
You don’t have to be most companies.
Please don’t be most companies.
Thanks for sticking with me through this entire series. I know it was dense. But the companies that commit to building this infrastructure right create compound advantages their competitors can’t match.
If you’re building this in your organization and want to talk through specific challenges, hit reply.
See ya next week,
Kaylee ✌
Incredible content, thx for sharing!